如何选择合适的试验设计类型(五)

当试验中涉及两个因素时,若两因素之间不存在交互作用或交互作用对观测结果的影响无统计学意义,并且,特定试验条件(通常为各因素不同水平的一种组合)下试验数据的测定误差在专业上允许的范围之内,此时,可以选用"随机区组设计、平衡不完全随机区组设计、具有一个重复测量的单因素设计、无重复试验双因素设计和两因素嵌套设计"五种设计之一.在上一期讲座中,我们向读者重点介绍无重复试验双因素设计和两因素嵌套设计这两种设计类型,重点阐述了两设计类型的定义、形式、特点、应用场合以及具体实施,并用实例加以论证.试验中所涉及到的两个因素都是试验因素,仅当有预试验表明两个试验因素之间的交互作用无统计学意义,且相同试验条件下定量观测指标的取值离散度很小时,此时,可采用无重复试验双因素设计;试验因素对观测结果的影响有主次之分,或者试验因素之间存在自然属性上的嵌套关系,上层因素对观测结果的影响大于下层因素时,采用两因素嵌套设计较为适宜.     

如何正确运用方差分析——嵌套设计定量资料一元方差分析与SAS实现

本文目的是介绍嵌套设计及其定量资料的方差分析与SAS实现。若某项试验研究存在以下两个特点之一,则可考虑选择嵌套设计安排试验：①因素之间存在自然属性上的嵌套关系;②依据专业知识,各因素对定量观测结果的影响存在主次之分。前述第一个特点意味着与受试对象有关联的因素具备分组再分组的条件;前述第二个特点意味着各因素的地位是不平等的。因此,对定量资料进行方差分析时,需采用可变误差均方的计算公式。本文基于4个实例并借助SAS软件,实现了嵌套设计定量资料一元方差分析,并对SAS输出结果作出解释。     

试验设计类型之概述

本文目的是全面介绍与试验设计类型有关的概念以及学习和掌握试验设计类型知识的意义。通过介绍"研究类型、设计类型、比较类型和资料类型"的概念、临床试验研究中常见试验设计类型的实例并从不同角度来划分设计类型的思路,读者能够比较全面和深刻地把握设计类型的要领。本文首次将多因素设计类型归纳为"伴随式多因素试验设计"与"统筹式多因素试验设计"两大类,极大地方便了广大科研工作者结合自己的科研实际,合理选定并规划出所需要的多因素试验设计类型。     

试验设计类型之可以考察全部交互作用的多因素设计:析因设计

本文目的是介绍一种可以考察各级交互作用的多因素设计类型,即析因设计。通过介绍两个实际例子,总结出析因设计的八个特点及优点、缺点。针对实际应用时可能面临的情况或问题,人们可能会对析因设计进行改造,由此产生了析因设计的三种变形,即"含区组因素的析因设计""分式析因设计"和"多因素非平衡组合试验"。最后一种情形,是实际工作者常会使用的,但在对定量评价指标进行差异性分析时需要对原先的分组进行合理拆分,再进行恰当的组合,构造出多种标准的设计类型,以利于分析者选择正确的统计分析方法。     

试验设计类型之可以考察部分交互作用的多因素设计:裂区设计

本文旨在介绍一种多因素试验设计方法,即裂区设计。从农业、工业、医学试验研究的实际出发,呈现了三个多因素试验的实例。进而总结出裂区设计具有的两个突出特点,即"一个完整试验被整批重复"和"试验因素施加于受试对象时存在先后顺序之分"。准确辨识一个多因素试验研究定量资料是否来自裂区设计,其意义在于是否需要选择裂区设计定量资料方差模型处理资料。     

如何选择合适的试验设计类型（四）

当试验中涉及两个因素时,若两因素之间不存在交互作用或交互作用对观测结果的影响无统计学意义,并且,特定试验条件（通常为各因素不同水平的一种组合）下试验数据的测定误差在专业上允许的范围之内,此时,可以选用“随机区组设计、平衡不完全随机区组设计、具有一个重复测量的单因素设计、无重复试验双因素设计和两因素嵌套设计”五种设计之一。在上一期讲座中,我们向读者重点介绍了随机区组设计、具有一个重复测量的单因素设计这两种设计类型,重点阐述了两设计类型的定义、形式、特点、应用场合以及具体实施,并用实例加以论证。     

试验设计类型之可以考察部分交互作用的多因素设计:正交设计与均匀设计

本文旨在介绍两种多因素试验设计方法,即正交设计与均匀设计。由于析因设计中的试验点数目过多,在既要较少试验点数目,又要能比较准确地揭示多因素对评价指标影响规律的要求之下,宜选择正交设计或均匀设计。正交设计具有"试验点均匀分散、整齐可比且可以事先安排少数交互作用项"的特点,因此,结果的可信度较高,只需采用方差分析处理定量资料;均匀设计具有"试验点极其均匀分散、正交性在一定程度上受到破坏,不能事先安排交互作用项"的特点,因此,允许很少数目的试验点,但需采用多重回归分析处理资料且分析结果不唯一。     

试验设计类型之无法考察交互作用的多因素设计:随机区组设计与拉丁方设计

本文目的是介绍两种无法考察交互作用的多因素设计类型,即随机区组设计与拉丁方设计。通过详细介绍与前述两种设计类型对应的"主要内容""操作方法"和"设计变形",全面展示了这两种设计类型的核心内容、要领和注意事项,为用户正确、合理地选用前述的两种设计类型创造了有利的条件。     

试验设计类型之单因素设计

本文目的是全面介绍与单因素设计有关的问题。通过详细分析单因素设计的几种亚型和由配对设计退化而成的单组设计,全面而深入地呈现出合理选用单因素设计的要领;通过对"组别"真相的揭示,提高读者识别"假单因素设计"的能力。真正的单因素设计只涉及一个专业上可以清楚命名的影响因素,其他任何非试验因素对评价指标的影响在该影响因素各水平组之间是均衡的,其与影响因素之间的交互作用在客观上是可以忽略不计的。     

试验设计类型之可以考察部分交互作用的多因素设计:重复测量设计

本文目的是介绍一种广泛应用于生物医药学和临床试验研究中的多因素试验设计类型,即重复测量设计。根据试验分组因素数目和重复测量因素数目的不同,此设计类型具有多种不同的亚型。此设计类型特别适用于需要考察接受不同处理的受试对象随着时间的推移、定量指标取值的动态变化规律。此设计类型的一个突出特点是定量指标在不同时间点之间具有不相等的相关性,在对其定量资料进行差异性分析时,需要采用多种不同的方差协方差结构的方差分析模型,并依据特定的判定标准,从中选择最合适的方差分析模型给出最终的分析结果。     

Status of single-case research designs for evidence-based practice

The single-case research design has become a paradoxical methodology in the applied sciences. While various experimental designs have been in place for over 50 years, there has not been wide acceptance of single-case methodology outside clinical and school psychology, or the field of special education. These methods were developed in the U.S.A., and their acceptance internationally has been even slower. The development and popularity of applied behavior analysis for persons with intellectual disabilities and the tremendously popular field of autism, and single-case designs have been very closely inter-related and have helped spread the use of single-case designs. The purpose of this paper is to review the history, types of designs, and their maturation in the applied fields of mental health and special education.     

Reliability of observers' subjective impressions of families: A generalizability theory approach

Abstract

Parenting was observed in videotaped interactions in 30 families referred for child conduct problems. Generalizability coefficients and the impact of varying numbers of raters were estimated. Two measurement designs were compared: All raters observed all families (“crossed” design) and a different rater observed each family (“nested” design). The crossed design provided higher generalizability coefficients than a nested design, implying inflated generalizability estimates if a crossed estimation model is used for a nested data collection. Three and four raters were needed to obtain generalizability coefficients in the .70–.80 range for monitoring and discipline, respectively. One rater was sufficient for a corresponding estimate for positive involvement and for an estimate in .80–.90 range for problem-solving. Estimates for skill encouragement were non-acceptable.     

Motion correction and the use of motion covariates in multiple-subject fMRI analysis

The impact of using motion estimates as covariates of no interest was examined in general linear modeling (GLM) of both block design and rapid event-related functional magnetic resonance imaging (fMRI) data. The purpose of motion correction is to identify and eliminate artifacts caused by task-correlated motion while maximizing sensitivity to true activations. To optimize this process, a combination of motion correction approaches was applied to data from 33 subjects performing both a block-design and an event-related fMRI experiment, including analysis: (1) without motion correction; (2) with motion correction alone; (3) with motion-corrected data and motion covariates included in the GLM; and (4) with non-motion-corrected data and motion covariates included in the GLM. Inclusion of covariates was found to be generally useful for increasing the sensitivity of GLM results in the analysis of event-related data. When motion parameters were included in the GLM for event-related data, it made little difference if motion correction was actually applied to the data. For the block design, inclusion of motion covariates had a deleterious impact on GLM sensitivity when even moderate correlation existed between motion and the experimental design. Based on these results, we present a general strategy for block designs, event-related designs, and hybrid designs to identify and eliminate probable motion artifacts while maximizing sensitivity to true activations.     

试验设计类型之无法考察交互作用的多因素设计:交叉设计

本文目的是介绍一种无法考察交互作用的多因素设计类型,即交叉设计。通过详细介绍与交叉设计类型对应的"主要内容""操作方法"和"设计变形",全面展示了这种设计类型的核心内容、要领和注意事项,为用户正确、合理地选用交叉设计类型创造了有利的条件。     

Partially nested designs in psychotherapy trials: A review of modeling developments

Objective: Individually-randomized psychotherapy trials are often partially nested. For instance, individuals assigned to a treatment arm may be clustered into therapy groups for purposes of treatment administration, whereas individuals assigned to a wait-list control are unclustered. The past several years have seen rapid expansion and investigation of methods for analyzing partially nested data. Yet partial nesting often remains ignored in psychotherapy trials. Methods: This review integrates and disseminates developments in the analysis of partially nested data that are particularly relevant for psychotherapy researchers. Results: First, we differentiate among alternative partially nested designs. Then, we present adaptations of multilevel model specifications that accommodate each design. Next, we address how moderation by treatment as well as mediation of the treatment effect can be investigated in partially nested designs. Model fitting results, annotated software syntax, and illustrative data sets are provided and key methodological issues are discussed. Conclusions: We emphasize that cluster-level variability in the treatment arm need not be considered a nuisance; it can be modeled to yield insights about the treatment process.     

Risk factors and study designs used in research of youths’ suicide behaviour—An epidemiological discussion with focus on level of evidence

Abstract

Introduction: Many different epidemiology study designs have been used to analyse risk factors for suicide behaviour. The purpose of this study was to obtain an insight into the current study design used in research on youths’ risk factors for suicide behaviour and to rank the studies according to level of evidence (LoE). Methods: We searched PubMed and psycINFO in order to identify relevant individual studies. Results: We included 36 studies of children and youth on suicidal behaviour and ideation—many rank low on LoE. For suicide, cohort design was often used, and mental illness (depression, substance abuse and severity of mental illness) was the most common risk factor. Cohort studies are ranked 2b, which is high according to LoE. For suicide attempts, survey was often used, and psychopathology, substance abuse and being exposed to suicidal behaviour were the most common risk factors. For suicidal ideation, survey was the only design used, and substance abuse and psychopathology the most common risk factors. Surveys are ranked 4, which are low according to LoE. Many risk factors were broad and unspecific, and standard definitions of outcome and exposure were rarely used. Conclusion: A good study of risk factors for suicidal behaviour would need a high LoE, as a high-powered longitudinal epidemiological study (cohort or case–control) of very specific risk factors. The factors would have high prevention potential, compared with more broad and unspecific risk factors, to which many people are exposed. We would recommend a cohort design (in high-risk populations) or a case–control design to identify risk factors, using clinical and/or register data instead of self-reported information, reporting adjusted estimates and using standard definition of suicidal outcome and risk factors.     

Decoding subjective mental states from FMRI activity patterns

In recent years, functional magnetic resonance imaging (fMRI) decoding has emerged as a powerful tool to read out detailed stimulus features from multi-voxel brain activity patterns. Moreover, the method has been extended to perform a primitive form of 'mind-reading,' by applying a decoder "objectively" trained using stimulus features to more "subjective" conditions. In this paper, we first introduce basic procedures for fMRI decoding based on machine learning techniques. Second, we discuss the source of information used for decoding, in particular, the possibility of extracting information from subvoxel neural structures. We next introduce two experimental designs for decoding subjective mental states: the "objective-to-subjective design" and the "subjective-to-subjective design." Then, we illustrate recent studies on the decoding of a variety of mental states, such as, attention, awareness, decision making, memory, and mental imagery. Finally, we discuss the challenges and new directions of fMRI decoding.     

Sample designs and sampling methods for the Collaborative Psychiatric Epidemiology Studies (CPES)

This paper provides an overview of the probability sample designs and sampling methods for the Collaborative Psychiatric Epidemiology Studies (CPES): the National Comorbidity Survey Replication (NCS-R), the National Study of American Life (NSAL) and the National Latino and Asian American Study of Mental Health (NLAAS). The multi-stage sample design and respondent selection procedures used in these three studies are based on the University of Michigan Survey Research Center's National Sample designs and operations. The paper begins with a general overview of these designs and procedures and then turns to a more detailed discussion of the adaptation of these general methods to the three specific study designs. The detailed discussions of the individual study samples focus on design characteristics and outcomes that are important to analysts of the CPES data sets and to researchers and statisticians who are planning future studies. The paper describes how the expected survey cost and error structure for each of these surveys influenced the original design of the samples and how actual field experience led to changes and adaptations to arrive at the final samples of each survey population.     

Optimal design for nonlinear estimation of the hemodynamic response function

Subject-specific hemodynamic response functions (HRFs) have been recommended to capture variation in the form of the hemodynamic response between subjects (Aguirre et al., [ 1998]: Neuroimage 8:360-369). The purpose of this article is to find optimal designs for estimation of subject-specific parameters for the double gamma HRF. As the double gamma function is a nonlinear function of its parameters, optimal design theory for nonlinear models is employed in this article. The double gamma function is linearized by a Taylor approximation and the maximin criterion is used to handle dependency of the D-optimal design on the expansion point of the Taylor approximation. A realistic range of double gamma HRF parameters is used for the expansion point of the Taylor approximation. Furthermore, a genetic algorithm (GA) (Kao et al., [ 2009]: Neuroimage 44:849-856) is applied to find locally optimal designs for the different expansion points and the maximin design chosen from the locally optimal designs is compared to maximin designs obtained by m-sequences, blocked designs, designs with constant interstimulus interval (ISI) and random event-related designs. The maximin design obtained by the GA is most efficient. Random event-related designs chosen from several generated designs and m-sequences have a high efficiency, while blocked designs and designs with a constant ISI have a low efficiency compared to the maximin GA design.